Context

The incorporation of high-resolution cameras into smartphones has allowed for a variety of medical applications including the use of lens attachments that provide telescopic, macroscopic, and dermatoscopic data, but the feasibility and performance characteristics of such a platform for use in dermatopathology have not been described.

Objective

To determine the diagnostic performance of a smartphone microscope compared to traditional light microscopy in dermatopathology specimens.

Design

A simple smartphone microscope constructed with a 3-mm ball lens was used to prospectively evaluate 1021 consecutive dermatopathology cases in a blinded fashion. Referred, consecutive specimens from the community were evaluated at a single university hospital. The performance characteristics of the smartphone platform were calculated by using conventional light microscopy as the gold standard. The sensitivity and specificity for the diagnosis of melanoma, nonmelanoma skin cancers, and other miscellaneous conditions by the phone microscopy platform, as compared with traditional light microscopy, were calculated.

Results

For basal cell carcinoma (n = 136), the sensitivity and specificity of smartphone microscopy were 95.6% and 98.1%, respectively. The sensitivity and specificity for squamous cell carcinoma (n = 94) were 89.4% and 97.3%, respectively. The lowest sensitivity was found in melanoma (n = 15) at 60%, although the specificity was high at 99.1%. The accuracy of diagnosis of inflammatory conditions and other neoplasms was variable.

Conclusions

Mobile phone–based microscopy has excellent performance characteristics for the inexpensive diagnosis of nonmelanoma skin cancers in a setting where a traditional microscope is not available.

Dermatologic disease accounts for substantial global morbidity and is a leading cause of nonfatal disability. It is the fourth leading cause of nonfatal global disease burden.1,2  Skin conditions are responsible for an estimated 582 to 609 disability-adjusted-life-years (a measure of disease burden quantified by years of life lost secondary to premature mortality + years lived with disability) per 100 000 in North America alone.1

It has a high prevalence and disease burden in underdeveloped countries.3  Skin biopsy and the pathologic information that it provides is foundational to the diagnosis and management of dermatologic disease. However, a lack of expertise in dermatopathologic-specimen interpretation in resource-poor areas may impede timely diagnosis and treatment.4,5

Several studies using modern high-resolution digital cameras equipped on mobile phones have demonstrated the feasibility of providing adjunctive histopathologic data for telemedical diagnostics in resource-poor health care settings.68  Mobile phone microscopy using a ball lens as a point-of-care test has been shown by Bogoch et al9  to have a high sensitivity for certain species of roundworm eggs in stool smears from Tanzanian children.

We investigated a modified version of this apparatus in the diagnosis of dermatopathology specimens and report several performance characteristics in both malignant and benign dermatologic conditions.

A mobile phone microscope was constructed by using a method modified from Smith et al8  and Bogoch et al.9  A 3-mm-diameter sapphire ball lens was purchased from Edmund Optics (part No. 43644; Barrington, New Jersey) for US $14. Initially, the apparatus holding the lens was constructed by using double-sided tape (3M, St Paul, Minnesota) similarly to Bogoch et al.9 However, our double-sided tape left a film on the ball lens as the apparatus was being scanned across the slides. Therefore, in lieu of tape, a clear segment of plastic was recovered from an unused binder backing, folded in half, and 2-mm-diameter hole was made with a 2-mm-diameter punch biopsy instrument. The ball lens was sandwiched between the plastic segments at the site of the 2-mm hole and stapled in place with office staples. It was then temporarily affixed to an iPhone 5 (Apple, Cupertino, California) directly over the camera by using tape (Figure 1). The phone microscope was held with one hand and then scanned over the specimen slide (Figure 2; Supplemental Video [see supplemental material file at http://dx.doi.org/10.5858/arpa.2014-0593-OA.s1 and www.archivesofpathology.org in the January 2016 table of contents]). The slides were backlit by a small, 55-lumen light-emitting diode flashlight (Performance Tools W2450, US$6; Jegs, Columbus, Ohio). The approximate magnification of the ball-lens system was ×66 (Figure 3, A and B). The Supplemental Video was captured directly from the iPhone 5 used in this study with the native Apple camera application on video mode while scanning a basal cell carcinoma. The video file was encoded by using MPEG-2 as an .avi file at 29 frames per second with a resolution of 480 × 270 pixels, with a duration of 41 seconds. This video was made for demonstration only and not reviewed as part of the study.

Figure 1.

The mobile phone microscopy apparatus with a 3-mm ball lens enveloped in a thin piece of plastic, stapled together, and taped to the phone over the camera sensor.

Figure 2.Phone microscopy slide reading. The technique for reading a slide using the ball-lens phone microscope with a single light-emitting diode (LED) flashlight providing illumination.

Figure 3.A, Nodular basal cell carcinoma: basaloid tumor cells with peripheral palisading and a cystic dilation viewed through conventional light microscopy. B, The same field as viewed through the 3-mm ball-lens microscope (hematoxylin-eosin, original magnifications ×100 [A] and approximate original magnification ×66 [B]).

Figure 1.

The mobile phone microscopy apparatus with a 3-mm ball lens enveloped in a thin piece of plastic, stapled together, and taped to the phone over the camera sensor.

Figure 2.Phone microscopy slide reading. The technique for reading a slide using the ball-lens phone microscope with a single light-emitting diode (LED) flashlight providing illumination.

Figure 3.A, Nodular basal cell carcinoma: basaloid tumor cells with peripheral palisading and a cystic dilation viewed through conventional light microscopy. B, The same field as viewed through the 3-mm ball-lens microscope (hematoxylin-eosin, original magnifications ×100 [A] and approximate original magnification ×66 [B]).

A total of 1130 consecutive dermatopathology specimens from mostly community dermatologists were evaluated. All excisional specimens (n = 101) were excluded because the diagnosis was previously established. Nail biopsy specimens (n = 8) were also excluded, as it had been previously noted that the smartphone-based platform could not resolve the appearance of fungal elements in the positive control. The smartphone microscopist was instructed to make all diagnoses based only on routine staining. Immunohistochemistry, special stains, or deeper sample levels were not made available to this microscopist in order to replicate field conditions. Margins were not determined on specimens examined by smartphone microscopy, nor were Breslow or Clark depths measured on any lesions. The smartphone microscopist had access to the same clinical information as the light microscopist, including patient sex, age, and clinical description if it was given. A single senior pathologist using a light microscope rendered a diagnosis for each slide, and a different pathologist using the smartphone platform rendered a diagnosis on the same slide. Ultimately, the results of 1021 biopsy specimens from routine light microscopy and mobile phone microscopy were analyzed. Diagnosis by traditional light microscopy was used as the gold standard to which the mobile-phone microscopy was compared. The microscopists were blinded to each other's results.

Excel 2007 (Microsoft, Redmond, Washington) was used to calculate sensitivities, specificities, and predictive values with 95% confidence intervals for basal cell carcinoma (BCC), squamous cell carcinoma (SCC), combined nonmelanoma skin cancers (NMSCs), and melanoma. The percentage correct for all other diagnoses was also calculated. This study was approved by the University of Texas Houston Medical School Institutional Review Board.

Among the 1021 specimens, 930 (91.1%) were shave biopsies and 91 (8.9%) were punch biopsies from 763 different patients. There were 136 BCCs, 94 SCCs, and 15 melanomas detected by light microscopy. Table 1 summarizes the sensitivity, specificity, and predictive values of the mobile-phone–based microscopy system for the diagnosis of BCC, SCC, combined NMSC, and melanoma. The sensitivity for BCC was the highest at 95.6% and lowest for melanoma at 60.0%.

Table 1.

Performance Characteristics of Mobile Phone–Based Microscopy for the Diagnosis of Malignant Cutaneous Lesions

The specificity was relatively high at 97.3%, 98.1%, and 99.1% for SCC, BCC, and melanomas, respectively. The specificity for NMSC overall was 97.6%. The misdiagnosed cases of BCC, SCC, and melanoma and their incorrect diagnoses are listed in Table 2. There were 2 other types of cutaneous malignancy in the 1021 specimens: a Merkel cell carcinoma that was misdiagnosed as a BCC by phone microscopy and a cutaneous B-cell lymphoma that was misclassified as spongiotic dermatitis. The light microscopist confirmed both of these diagnoses with immunohistochemistry, which the mobile phone microscopist did not use.

Table 2.

Discrepant Diagnoses Between Conventional Light and Mobile Phone Microscopy Among Basal Cell Carcinomas (BCCs), Squamous Cell Carcinomas (SCCs), and Melanomas

Diagnostic categories for all other specimens are listed in Table 3 along with their percentage correct. Excluding molluscum and verruca, there were 2 other biopsy specimens with an infectious etiology, one of which was dermatophytosis interpreted as “basketweave orthokeratosis with parakeratosis and superficial perivascular lymphocytic infiltrate” and phaeohyphomycosis called SCC. The dermatophyte was later confirmed by periodic acid–Schiff stain ordered and reviewed by the light microscopist.

Table 3.

Diagnostic Categories for the 1021 Dermatopathology Specimens Examined and Their Percentage Correct as Compared to Light Microscopy

There were 80 specimens categorized as inflammatory, of which 59 (73.8%) were diagnosed correctly with smartphone microscopy (Table 4). The average phone microscopy diagnosis rate was approximately 18 to 20 specimens every 30 minutes, while the light microscopist averaged 15 minutes for that number.

Table 4.

Ten Most Frequent Inflammatory Diagnoses Among the 80 Inflammatory Specimens

Mobile teledermatology can bring much-needed expertise to resource-poor health care settings. Although slight discrepancies exist, teledermatologic diagnoses generally have good diagnostic concordance with in-person visits.1013  Additional information in the form of dermatopathologic data has been shown to increase diagnostic accuracy of teledermatology cases.4  In remote areas without access to pathology services, but where histologic specimen processing is available, telepathology from remote sites using store and forward images, remote robotic-controlled microscopes, or mobile phones attached to conventional microscopes have been shown to be both feasible and diagnostically helpful.6,7,1416  One study examined the use of a camera-mounted microscope to stream video to a dermatopathologist viewing the slide images on a tablet computer and found a concordance rate of 97.6% to conventional light microscopy.17  However, the test performance characteristics of this system in their 93 specimens were not evaluated.

This study prospectively evaluated a large number of dermatopathology specimens, using a ball-lens microscope apparatus on a smartphone, and determined the sensitivity, specificity, and predictive values of this low-cost system. The ball lens is economical at US $14. Smartphones, whether Android or iOS operating system based with similar optical and processing capabilities, can be purchased for less than US$150. The apparatus can be assembled with basic office supplies, is nonbulky, and is highly portable, all of which makes it convenient for use in resource-poor settings. Microscopy using smartphones offers additional capabilities such as store-and-forward imaging, live-streaming of video, and the development of software enabling complex image analyses.16

Although in a prototype form, this microscopy system had a high degree of sensitivity and specificity for the diagnosis of NMSC and performed well for several different categories of neoplastic and inflammatory lesions (Table 3). The diagnostic category of benign epidermal lesions had a poor concordance rate of 69.8% when compared to conventional light microscopy. The sensitivity for diagnosis of melanoma was similarly poor at 60%. The differentiation between epidermal lesions such as solar lentigo and actinic, lichenoid, or seborrheic keratosis may require a resolution that is not attainable with a 3-mm ball lens. The same holds true for the pagetoid scatter, lentiginous growth, and melanocytic atypia that help distinguish melanoma from other entities. One possible solution is the use of a smaller 1- or 2-mm ball lens, which would increase the magnification at the expense of field of view or the use of a lens-free mobile microscopy platform.16

Owing to time constraints, each pathologist only reviewed specimens on their respective platforms. Because multiple pathologists on each platform were not employed, estimations of interobserver variability were precluded and comparison to other telemedicine or telepathology studies using this design is made more difficult. We propose that, although there are 2 possible sources for a misclassification bias—one due to technologic limitation of the smartphone apparatus and the other being actual differences in the interpretation of the histologic findings—the high diagnostic concordance between many of the biopsy categories, with more than 1000 cases examined, suggests the platform holds promise. Furthermore, given the potential for interobserver variability, there is a theoretical possibility that the smartphone microscopist may have performed better than the light microscopist, at least for some diagnoses. We believe this scenario to be less likely given the resolution limitations of the smartphone apparatus in comparison to the light microscope, and because the light microscopist assigned as the gold standard was a senior faculty pathologist with more than 30 years of experience. In future studies, the more challenging diagnostic categories for the smartphone platform, such as inflammatory disorders, melanoma, and benign epidermal lesions, may require the participation of multiple pathologists and measurement of interobserver variability.

Although the specimens were evaluated at a single academic institution, they were received from dermatologists both in academic and community settings. The relative lack of infectious disease biopsy specimens precludes proper evaluation of this category. Also, nails, special stains, immunostains, and excisional specimens were not evaluated in order to replicate basic field conditions; however, future iterations of this platform could be optimized for these specimens.16  In its current form, with a 93% sensitivity and 97.7% specificity for NMSC, the 3-mm ball-lens apparatus could be used to help triage potential surgical candidates identified in remote field locations, especially if combined with in-field histologic processing.18

Our study demonstrates the potential for a high-performing, low-cost smartphone microscopy system in the diagnosis of cutaneous disease. Studies with larger sample sizes of pigmented lesions, NMSC subtypes, and dermatologic infections could help improve this modality for point-of-care diagnostics in resource-poor settings.

The authors are indebted to Megan Abuzeid, MD, for assistance with slide collection and organization.

1
Hay
RJ
,
Johns
NE
,
Williams
HC
,
et al
.
The global burden of skin disease in 2010: an analysis of the prevalence and impact of skin conditions
.
J Invest Dermatol
.
2014
;
134
(
6
):
1527
1534
.
2
Murray
CJ
,
Vos
T
,
Lozano
R
,
et al
.
Disability-adjusted life years (DALYs) for 291 diseases and injuries in 21 regions, 1990-2010: a systematic analysis for the Global Burden of Disease Study 2010
.
Lancet
.
2012
;
380
(
9859
):
2197
223
.
3
Hay
RJ
,
Fuller
LC.
Assessment of dermatological needs in resource-poor regions
.
Int J Dermatol
.
2011
;
50
(
5
):
552
557
.
4
Tsang
MW
,
Kovarik
CL.
The role of dermatopathology in conjunction with teledermatology in resource-limited settings: lessons from the African Teledermatology Project
.
Int J Dermatol
.
2011
;
50
(
2
):
150
156
.
5
Tsang
MW
,
Kovarik
CL.
Global access to dermatopathology services: physician survey of availability and needs in sub-Saharan Africa
.
.
2010
;
63
(
2
):
346
348
.
6
Tuijn
CJ
,
Boefman
BJ
,
van Beijma
H
,
Oskam
L
,
Chevrollier
N.
Data and image transfer using mobile phones to strengthen microscopy-based diagnostic services in low and middle income country laboratories
.
PLoS One
.
2011
;
6
(
12
):
e28348
.
7
Zhu
H
,
Yaglidere
O
,
Su
TW
,
Tseng
D
,
Ozcan
A.
Wide-field fluorescent microscopy on a cell-phone
.
Conf Proc IEEE Eng Med Biol Soc
.
2011
;
2011
:
6801
6804
.
8
Smith
ZJ
,
Chu
K
,
Espenson
AR
,
et al
.
Cell-phone-based platform for biomedical device development and education applications
.
PLoS One
.
2011
;
6
(
3
):
e17150
.
9
Bogoch
II
,
Andrews
JR
,
Speich
B
,
et al
.
Mobile phone microscopy for the diagnosis of soil-transmitted helminth infections: a proof-of-concept study
.
Am J Trop Med Hyg
.
2013
;
88
(
4
):
626
629
.
10
Warshaw
EM
,
Lederle
FA
,
Grill
JP
,
et al
.
Accuracy of teledermatology for pigmented neoplasms
.
.
2009
;
61
(
5
):
753
765
.
11
Warshaw
EM
,
Lederle
FA
,
Grill
JP
,
et al
.
Accuracy of teledermatology for nonpigmented neoplasms
.
.
2009
;
60
(
4
):
579
588
.
12
Lamel
SA
,
Haldeman
KM
,
Ely
H
,
Kovarik
CL
,
Pak
H
,
Armstrong
AW.
Application of mobile teledermatology for skin cancer screening
.
.
2012
;
67
(
4
):
576
581
.
13
Wolf
JA
,
Moreau
JF
,
Akilov
O
,
et al
.
Diagnostic inaccuracy of smartphone applications for melanoma detection
.
JAMA Dermatol
.
2013
;
149
(
4
):
422
426
.
14
Bellina
L
,
Missoni
E.
Mobile cell-phones (M-phones) in telemicroscopy: increasing connectivity of isolated laboratories
.
Diagn Pathol
.
2009
;
4
:
19
.
15
Terry M.
Telemicroscopes
and point-of-care diagnostics team up with smartphones
.
Telemed J E Health
.
2011
;
17
(
5
):
320
323
.
16
Zhu
H
,
Isikman
SO
,
Mudanyali
O
,
Greenbaum
A
,
Ozcan
A.
Optical imaging techniques for point-of-care diagnostics
.
Lab Chip
.
2013
;
13
(
1
):
51
67
.
17
Speiser
JJ
,
Hughes
I
,
Mehta
V
,
Wojcik
EM
,
Hutchens
KA.
Mobile teledermatopathology: using a tablet PC as a novel and cost-efficient method to remotely diagnose dermatopathology cases
.
Am J Dermatopathol
.
2014
;
36
(
1
):
54
57
.
18
Calhoun
K
,
Lin
A
,
Bryant-Greenwood
P
,
Lum
C
,
Johnson
D
,
Namiki
T.
Field histology: point-of-care microscopic technique
.
Arch Pathol Lab Med
.
2011
;
135
(
2
):
207
210
.

## Author notes

The authors have no relevant financial interest in the products or companies described in this article.